Skip to content

CHANGELOG.md

Latest commit

 

History

History
59 lines (41 loc) · 2.85 KB

CHANGELOG.md

File metadata and controls

59 lines (41 loc) · 2.85 KB
Raw

Changelog

[0.1.2] — 2025-03-26

Changed

  • BOM fast path: UTF-8 with BOM now reports encoding="utf_8_sig" (aligned with codecs / CandidateSelector), not utf_8.
  • Streaming: In-band HTML/XML hints are probed on every feed() until a hint is found (scan limited to the first 4KB inside _probe_inband_hint); shared regex patterns imported from hints.py; codecs imported at module level.
  • detect_multi: Adjacent same-encoding merges no longer re-run from_bytes on the full merged span; byte_count on the shared DetectionResult is updated with dataclasses.replace.
  • repair_bytes: Explicit max_iterations and chains parameters instead of **kwargs: object.
  • Coverage: fail_under raised from 50 to 75 (current suite ~75%; 85% remains a stretch goal with more api / fingerprint tests).

Fixed

  • DocumentSegment / MultiEncodingResult now use slots on Python 3.10+ like other dataclasses in the package.

[0.1.1] — 2025-03-26

Added

  • examples/ scripts (basic detection, detect(), streaming, repair, HTTP hints, multi-encoding)
  • MkDocs documentation site and GitHub Actions workflow Docs Pages → GitHub Pages
  • [project.optional-dependencies] group docs (mkdocs, mkdocs-material, pymdown-extensions)
  • project.urls.Documentation in pyproject.toml

Fixed

  • README logo on PyPI: use raw.githubusercontent.com URL (sdist has no assets/ for the image)
  • Source distribution: include LICENSE in maturin include so PyPI accepts License-File metadata

[0.1.0] — 2025-03-26

First published release.

Added

  • from_bytes(), from_path(), from_fp(), is_binary() API
  • detect() chardet / charset-normalizer drop-in compatibility
  • detect_stream() for iterator-based streaming detection
  • StreamDetector for incremental detection; snapshot(), is_stable, auto-stop when encoding is stable (configurable)
  • In-band hints from HTML meta tags and XML declarations in StreamDetector
  • Byte-distribution fingerprinting with pre-computed lookup table
  • repair() and repair_bytes() for mojibake detection and repair; is_mojibake(); RepairResult
  • hint_from_http_headers(), hint_from_content(), best_hint() for standalone hint extraction
  • detect_multi() for multi-encoding documents; MultiEncodingResult and DocumentSegment
  • Optional Rust core extension (pip install "bytesense[fast]")
  • CLI: bytesense <file>
  • py.typed marker for PEP 561 compliance
  • Full type annotations, mypy strict mode
  • GitHub Actions CI across 3 OS × multiple Python versions

Changed

  • LANGUAGE_ENCODINGS aligned with languages in CHAR_FREQUENCIES
  • Candidate shortlist tuned for speed while preserving accuracy targets

Fixed

  • Benchmark and packaging fixes (e.g. UTF-8 BOM test fixture, build backend configuration)